Study On Hazy Image Enhancement In Motion Blurring Condition

Due to the scattering of atmospheric particles in recent years, visibility of roads captured by visual system suffer from poor contrasts and colors under fog conditions. This weather brings a lot of inconvenience no matter in military, transportation, or daily life. On the haze and dust grounds, pictures taken are not clear enough to judge. Moreover, images tend to become a blur due to the shake of hands and installations. Consequently, it is essential to tackle these motioning-vague pictures both in manufacturing and our life, which become a hot issue in handling digital-figure of computer-vision sphere.Removing the haze belongs to image enhancement. And wipe of the motion blur belongs to image recovering. This paper finishes the four tasks as followed:(1) Concerning the part of blurring image, the concept of point spread function is advanced. This paper proposes a method of identifying the blurred length in the spatial domain on the basis of image differential coefficient and autocorrelation function. The paper classifies the motion-blurred images or motion blurred images rotating from any anglesto the horizontal level into different levels according to their sizes, calculates the horizontal differential coefficients of each layer image and the autocorrelation function of horizontal level, adds up the sum of each column, draws the relevant differential curve, finds the minimum distance used to calculate the blurred length, estimates the image at this layer from the differential curves. Finally, the paper selects a group of similar length estimated for the average of blurred length of the whole image. The result of the experiment shows that segmental identification proposed in this paper functions more efficiently and more accurately. (2) According to removing haze of images, the standard algorithms of Retinex and scattering of atmospheric particles are studied. This paper analyses their shortages and improved them respectively. It proposed a novel image enhancement algorithm by improving the standard multiscale Retinex algorithm. A modified Retinex output using a combination of two logarithmic functions enhances the details in highlights as well as the details in shadows. Its name is mlog(-) function. The algorithm conversed RGB color space to HSI space. Hue was kept unchanged. The function based on center self-adaptive adjustment was used to enhance overall brightness of image. And linear stretching was used to adjust saturation. (3) This paper studies the scattering of atmospheric particles algorithm. Secondly, it shows a few defects of the primary algorithm. Thirdly, it improves the algorithm. This paper proposes novel fast defogging method from a single image of a scene based on a fast bilateral filtering approach. Using atmospheric transportation veil and hue mapping are to result in n a large improvement in image visibility. (4) Regarding to this paper task that is motion blurring image in hazy weather, it sets up the system to enhance the images.Theoretical analysis and experimental results on real images demonstrate that, methods used in this paper are effective for clearness of fog-degraded images, and can be applied to the practical situations. Also, this research will be inspiration for the further investigations on the process of images taken in other bad weather.